EP2989787A1

EP2989787A1 - Image capture with addition of accumulation signals for adjacent photodetectors

Info

Publication number: EP2989787A1
Application number: EP14726685.2A
Authority: EP
Inventors: Philippe Luquet
Original assignee: Airbus Defence and Space SAS
Current assignee: Airbus Defence and Space SAS
Priority date: 2013-04-25
Filing date: 2014-04-22
Publication date: 2016-03-02
Also published as: CA2909900A1; CN105264881B; US20160073045A1; US9554065B2; FR3005226B1; FR3005226A1; CN105264881A; IL242174B; WO2014174195A1

Abstract

An image capture instrument implements an addition of accumulation signals for photodetectors (10) which are adjacent within a row (1) of photodetectors. Segments (S1, S2, S3,…) are defined in the row with one and the same common length of segment (L), while being progressively offset by an offset length (M) which is constant. The accumulation signals are added together for photodetectors which belong to one and the same segment. The offset length is less than the segment length, and greater than half said segment length. Artifacts in the image captured, which are due to spectrum aliasing, are thus decreased. The signal-to-noise ratio of the image is increased.

Description

IMAGE ENTRY WITH ADDITION OF ACCUMULATION SIGNALS FOR ADJACENT PHOTODETECTORS

The present invention relates to an image pickup instrument in which accumulation signals produced by adjacent photodetectors are added together. It also relates to a method that is implemented in such an instrument. The aim of the invention is to improve the quality of the images that are captured when the conditions of formation and image detection generate a spectral aliasing, or when the signal-to-noise ratio of the image is insufficient.

A feature of an image capture instrument is its modulation transfer function, or FTM. This function characterizes the ability of the instrument to render in the images that are captured, patterns of the photographed scene that have varying dimensions. Thus, the modulation transfer function is the contrast ratio between the image and the scene, for a pattern whose intensity varies sinusoidally and which is identified by its spatial frequency in the image. The modulation transfer function is obtained by varying the spatial frequency of the pattern, and determining its contrast in the image as a function of this spatial frequency.

In known manner, the capture of an image by the instrument has a low-pass filter effect. This effect appears as a decay of the modulation transfer function as the spatial frequency of the pattern in the image increases.

Several contributions together constitute the modulation transfer function, in the form of a product of these contributions to constitute the modulation transfer function of the complete instrument. A first contribution is produced by the imaging optics that is used in the instrument. This first contribution is a decreasing function of the spatial frequency in the image, and is substantially zero for spatial frequencies which are greater than a cutoff frequency fc. This frequency is given by the following formula: fc = 1 / (N ^■ λ), where: λ is a wavelength of the radiation which is used to form the image, and which comes from the imaged scene by and N is the numerical aperture of the imaging optics: N = f / D, where f and D are respectively a focal distance and a pupil diameter of this optic.

Thus, the spectral components of the scene that correspond to spatial frequencies in the image that are greater than the cutoff frequency f _c , are removed from the image by the optical itself. In other words, the image of the scene formed by the optics contains only spectral components whose spatial frequencies are lower than the cutoff frequency fc.

A second contribution to the modulation transfer function is produced by the image sensor that is used in the instrument. This sensor is located in the focal plane of the imaging optics, and comprises at least one row of photodetectors which have identical individual dimensions and are aligned with a pitch p which is constant along the array.

In known manner, such a sensor performs a sampling of the image, with a sampling frequency ÎE which is equal to 1 / p, when an accumulation signal is read separately in each photodetector to constitute a different point of the image. captured image.

When the sampling frequency f _E by the photodetectors is less than twice the cutoff frequency f _c , a spectrum fold occurs because of the insufficient sampling level. This spectrum folding concerns the spectral components of the scene whose spatial frequencies in the image are between f _E / 2 and f _c . The intensities of these spectral components appear added to those whose spatial frequencies are lower than f _E. The limit of half of the sampling frequency f _E to the sensor's ability to provide a true representation of the spectral components of the image is called the Nyquist frequency. When the image sensor is assumed to be perfect, the value of its contribution to the modulation transfer function, for the spatial frequency in the image which is equal to half of the sampling frequency f _E , is equal to sinc (n / 2) "0.64, when each photodetector separately produces an accumulation signal that corresponds to a different image point. In the preceding formula, sinc (x) is the sine-cardinal function, or [sin (x)] / x when applied to an argument x.

The modulation transfer function that results from the image forming optics and the image sensor then has a value that is still significant for half the sampling frequency f _E. The spectrum folding then produces artifacts in the captured image, which are visible especially when this image is merged with another of the same scene but whose sampling frequency is higher.

One solution for reducing these artifacts is to reduce the size of the photodetectors to reduce their pitch p in the sensor array. In this way, the sampling of the image is increased. But the image sensor, with photodetectors that are smaller, is more complex. The data rates to be processed are then greater, which also leads to increased complexity for the on-board data processing chain. In addition, the image sensor that is used can be imposed by various constraints, including other uses of the same sensor when a multipurpose mission is provided. In other words, the individual dimension and the pitch of the photodetectors can be imposed.

Another solution for reducing artifacts due to aliasing is to reduce the pupil size of the imaging optics. Thus, the value of the modulation transfer function at half of the sampling frequency is reduced. In this way, the parasite contribution of the folded frequencies, in the contrasts of the spectral components of the image whose spatial frequencies are lower than the sampling frequency, is reduced. But, imaging optics can be common to many functions. In particular, this optics can be common to several image acquisition channels that are used simultaneously, for example with intervals that are different for the wavelength of the radiation. The size of the pupil of the imaging optics is then fixed, without being able to be adapted separately according to the sensor of each imaging channel. Finally, it is also known to modify the effective dimension and the effective pitch of photodetectors for images that are captured by "agglomerating" photodetectors which are adjacent along a row of the sensor. For this, the photodetectors are grouped into row segments, successive along the row and disjoint. The term "agglomeration of neighboring photodetectors", or "photodetector binning" in English, a reading mode of the photodetector accumulation signals according to which the respective signals of photodetectors belonging to the same segment are added together. Figure 3 illustrates such a reading mode of the image sensor, as known before the present invention. In this figure, the reference 10 denotes the individual photodetectors of the image sensor 1, and S1 ', S2', S3 ', S4' ... denote successive and disjoint grouping pairs of the photodetectors 10. Consequences of a such agglomeration of the photodetectors are in particular:

the spatial sampling frequency of the image is halved, which represents a reduction in imaging performance; and

the signal-to-noise ratio relative to each intensity value that is used in the image as it is input is improved.

Such agglomeration of the photodetectors, which aims to increase the signal-to-noise ratio by accepting resolution degradation, also causes or aggravates image artifacts that are caused by spectrum folding.

From this situation, a first object of the invention is to more effectively reduce the artifacts that are produced by the aliasing of the captured images. A second object of the invention is to improve the signal-to-noise ratio for each intensity value that constitutes the captured image. Finally, complementary objects of the invention are to avoid making the image capture instrument and / or the processing of the images that are captured as complex as possible, as well as to avoid increasing the cost price of the image capture instrument. instrument. In particular, an additional object of the invention is to avoid increasing the sampling frequency of the image for a cutoff frequency of the imaging optics which remains constant.

To achieve at least one of these or other objects, the present invention provides an image capture instrument that includes:

an image forming optics and a row of photodetectors as previously described; and

a controller adapted to control readings of the accumulation signals produced respectively by the photodetectors.

The controller is further adapted to define a series of segments within the array, each of which includes the same number of adjacent photodetectors, referred to as segment length, and to control the readouts of the accumulation signals by summing those of which photodetectors belong to the same segments. In other words, the controller is adapted to achieve an agglomeration ("binning") of the photodetectors.

The controller is also adapted to define the segments so that an offset length M between two successive segments in the array, expressed as a number of photodetectors, is constant along the array, and greater than 1 / (2 ^■ f _c ^■ p). In this way, the instrument has a new IE sampling frequency of the image that is formed by the optics, equal to 1 / (M ^■ p). This new sampling frequency ÎE is less than twice the cutoff frequency f _c . The instrument of the invention therefore operates under conditions such that spectrum folding occurs. These features imply that the offset length M is non-zero.

The instrument of the invention also has the following additional feature: the segment length is strictly greater than the offset length M, and strictly less than twice this same offset length M.

This additional feature results in a reduction in the contribution of the image sensor to the modulation transfer function (MTF) of the instrument. But a reduction of the FTM is contrary to what is generally sought by the specialist in space optics, the aim of which is rather to increase it when it aims at an improvement of the image quality. However, by reducing the contribution of the image sensor to the MTF, the parasitic contribution of the spectral components of the scene whose spatial frequencies are folded in the image that is captured, is decreased. The image that is captured by agglomerating the photodetectors according to the invention therefore has fewer artifacts caused by the aliasing of the spectrum, which would appear in particular when this image is fused with another image whose sampling frequency would be higher. As a consequence of the invention, two successive segments of photodetectors which are used to add the accumulation signals during the reading, have between them an overlap of at least one photodetector. In addition, this overlap is smaller than half the length of each segment. In addition, the addition of the accumulation signals for each segment, which produces each intensity value of the captured image, has an improved value for the signal-to-noise ratio. Indeed, the total area of radiation collection that corresponds to this intensity value, is proportional to the segment length that is used to apply the invention. The invention also provides an image capture method, which comprises the following steps:

- provide imaging optics;

arranging the row of photodetectors in the focal plane of the imaging optics; defining the series of segments inside the row, each of which comprises the same number of adjacent photodetectors, called segment length; and

control the reading of the accumulation signals which are produced respectively by the photodetectors, by adding those whose photodetectors belong to one and the same segments.

In the method of the invention, a shift length M between two successive segments in the array, expressed as a number of photodetectors, is constant along the row, and greater than 1 / (2 - f _c - p).

The method of the invention is characterized in that the segment length is strictly greater than the offset length M, and simultaneously strictly less than twice this offset length M. However, the contrasts of the spectral components of the image of which the spatial frequencies are between the cutoff frequency of the imaging optics and half of the latter, are low in practice because of the values of the instrument's modulation transfer function for these spatial frequencies . In this case, and for preferred embodiments of the invention methods, it may be advantageous that the shift length M between two successive segments in the row is greater than 1 / (f _c ^■ p). The overlap between two successive segments is thus smaller. The sampling spatial frequency f _E of the image that is captured is then less than the cutoff frequency f _c , in addition to being less than twice the fc. The image sensor that is selected for the instrument can thus be less expensive, and the amount of image data is reduced.

In particular, the shift length M between two successive segments in the row, can be between 2 / (f _c p ^■) and 10 / (f _c ^■ p). In this way, the cutoff spatial frequency fc may be between two and ten times the sampling spatial frequency EE.

For example, the segment length may be equal to three adjacent photodetectors, and the offset length M between two segments successive ones in the row may be equal to two adjacent photodetectors.

In general, the photodetectors which are concerned by the invention may be elementary photodetectors of an image sensor, or radiation detection units which already result from an agglomeration of several elementary photodetectors of the image sensor. . The term photodetector which is used in the present description covers all these two modes of use of an image sensor.

In preferred implementations of the method of the invention, the addition of the accumulation signals for the photodetectors belonging to one and the same segments when the segment length is strictly greater than the offset length M and strictly less than the double of the same offset length M, may correspond to a first reading mode which is available among several alternative reading modes. Then, another series of segments can be defined inside the row of the photodetectors, for another of the reading modes, with the same length of offset M between two successive segments as in the first reading mode, but with a new segment length which is equal to the offset length M. Thus, two segments which are successive in the row of photodetectors no longer overlap with each other in the other reading mode. The accumulation signals are added in this other reading mode, for photodetectors belonging to one of the segments of the other series. In this case, the method may comprise an initial step of selecting one of the reading modes.

Within an image capture instrument which is in accordance with the invention, the two alternative reading modes can be defined and each controlled by the controller. The image capture instrument is then versatile, and can be adapted in this way for several missions without hardware modification.

For some applications of the invention in the field of spatial imaging, the image forming optics and the array of photodetectors may be parts of an image pickup instrument that is mounted onboard a satellite. The process can then be implemented to capture at least an image of a surface portion of the Earth.

Other features and advantages of the present invention will appear in the following description of nonlimiting exemplary embodiments, with reference to the appended drawings, in which: FIG. 1 represents an image acquisition instrument to which the present invention can be applied;

FIG. 2 illustrates a reading mode by agglomeration of photodetectors, as proposed by the present invention; and

FIG. 3 illustrates another mode of reading by agglomeration of photodetectors, known before the present invention.

For the sake of clarity, the dimensions of the elements which are represented in these figures do not correspond to real dimensions nor to actual dimension ratios. In addition, identical references which are indicated in different figures designate identical elements or which have identical functions.

In accordance with Figure 1, an image pickup instrument includes an image forming optics 100, an image sensor 1 and an image capture controller 2. The controller 2 is denoted CTRL and is generally a dedicated electronic unit, separated or not from the sensor 1. The imaging optics 100 may be of any type, based on lenses or mirrors. In particular, it can be adapted for space imaging missions. In this case, the optical 100 may be a telescope of a configuration known to those skilled in the art, such as a telescope with two mirrors. References 101 and 102 then designate respectively the primary mirror and the secondary mirror of the telescope. In known manner, the opening of the beams of radiation that enter the optics 100 is limited by the pupil P of this optic, whose diameter is denoted D. In the case of a telescope, the entrance pupil corresponds to the more often at the peripheral boundary of the primary mirror 101. The imaging optic 100 has a cutoff frequency f _c which is effective for the spectral components of the formed image, function of the spatial frequency of these components. The expression of this cut-off frequency f _c has been recalled at the beginning of the present description.

The photosensitive surface of the image sensor 1 is located in the imaging plane 103 of the optics 100, which most often corresponds to the focal plane thereof. Thus, radiation beams F1 and F2 which come from different points of a scene which is situated in the input field of optics 100, converge at separate points of the photosensitive surface of the sensor 1.

The image sensor 1 can be of any type: matrix sensor, online sensor or time-shifted integration sensor, designated by the acronym TDI for "time delay integration". The sensor 1 then comprises at least one row of photodetectors 10, which are juxtaposed in a row direction denoted R. The row may contain 3500 photodetectors 10, for example. To capture an image, each photodetector 10 individually produces an accumulation signal which is denoted I, and which is proportional to a quantity of radiation received by this photodetector during a given exposure time. In the following, but not limited to, we can consider that the sensor 1 is an in-line sensor, that is to say consisting of a single row of photodetectors. Finally, the controller 2 controls the operation of the sensor 1 to capture each image. In addition to the step of exposing each photodetector 10, it controls a step of reading the accumulation signals I. The image sensor 1 and the controller 2 can be adapted to capture the images according to different modes of operation, such as that the simultaneous exposure of all photodetectors 10, designated by "snapshot" in English, or a continuous reading mode of photodetectors, designated "rolling" in English. But preferably, the image sensor 1 and the controller 2 are adapted to combine the invention with a "pushbroom" type image capture mode. Those skilled in the art will understand that the invention which is described hereinafter can be combined with a wide variety of modes of operation for the image sensor 1. The controller 2 outputs an image, which is constructed by assigning intensity values respectively to points discrete of an image frame.

In a reference method for image capture, each point of the image frame may correspond to only one of the photodetectors 10, so that the respective intensity values of the points of the captured image correspond to one-to-one. one to the accumulation signals I which are read separately in the photodetectors 10. The imaging characteristics which are then obtained depend on those of the imaging optics 100 and the image sensor 1. These features include:

the cutoff frequency fc of the optics 100; the individual dimension d of photodetectors 10 according to the direction R.

More precisely, d is the size of the photosensitive surface portion of the image sensor 1, which corresponds to a single photodetector; and

the step p of the photodetectors 10 in the direction R, defined as the distance in this direction between two edges of the same side, relative to two photodetectors 10 which are neighbors within the row. In other words, the pitch p is the spatial period of the photodetectors 10 in the photosensitive surface of the sensor 1, according to the direction R.

The gap between pitch p and dimension d is the width of a dead band that exists between two adjacent photodetectors in the row. This difference is neglected in the following, but the skilled person will take into account if necessary to read the explanations given below.

The exact numerical values of these characteristics do not matter with respect to the principle of the invention, and depend on the actual components that are selected to constitute the image capture instrument. Likewise, the dimension e of the photodetectors 10, measured perpendicularly to the direction R, does not matter with respect to the invention.

The image capture methods which are in accordance with the invention use an agglomeration of the photodetectors 10. Some of the imaging features presented, including the contribution of the sensor 1 to the modulation transfer function and sampling frequency, then further depend on the agglomeration pattern that is used.

However, the contribution of the imaging optics 100 to the modulation transfer function of the instrument, referred to as the first contribution above, is not changed. As a result, the cutoff frequency fc of the optics 100 is unchanged.

Figure 2 illustrates a particular implementation of the invention. The photodetectors 10 are grouped into continuous segments in the direction R, with a segment length L which is common for all the segments, and an offset length M between two successive segments in the direction R, which is also constant. The successive segments are denoted S1, S2, S3, S4 ..., for example from one end of the row of the photodetectors 10. The lengths M and L are obviously each greater than or equal to unity, it is ie each larger than a single photodetector.

The invention introduces the following condition: M <L <2 ^■ M. In other words, the offset length M is less than the segment length L, and greater than half of this same segment length L.

The first inequality (M <L) expresses that two successive segments in the array overlap with each other, i.e. they have at least one photodetector 10 in common.

The second inequality (M> L / 2) expresses that this overlap between two successive segments is shorter than half the length of each segment. In the example shown in FIG. 2, the length L of each segment S1, S2, S3, S4... Is equal to three times the pitch p, corresponding to segments which each consist of three successive photodetectors 10. . The offset length M is equal to twice the pitch p, corresponding to successive segments which are offset from each other by two successive photodetectors 10. As a result, two successive segments have an overlap which has the length L - M in the direction R, ie an overlap of a photodetector 10 in the example shown. The controller 2 then performs an addition of the accumulation signals I for the photodetectors 10 belonging to one of the segments S1, S2, S3, S4. Thus, a first intensity value is obtained from the sum of the respective accumulation signals I of the three photodetectors 10 of the segment S1, a second intensity value is obtained from the sum of the respective accumulation signals I of the three photodetectors 10 of the segment S2, a third intensity value l ₃ is obtained from the sum of the respective accumulation signals I of the three photodetectors 10 of the segment S3, etc. Such an addition of the accumulation signals I within each segment S1, S2, S3, S4... Can be carried out in any way accessible to those skilled in the art: in a hardware way at the level of the circuits for reading the sensor 1, in a software manner within the controller 2, on analog signals, but also on digital signals. The image that is output by the controller 2 is constructed by assigning the intensity values, l ₂ , I ₃ . . . at different points of the image frame.

The length of the shift then replaces the step p of the individual photodetectors 10 in the row 1, for the sampling which is thus carried out inside the image formed by the optics 100. In other words, the sampling frequency of the instrument is then: f _E = 1 / (M ^■ p). In the particular example of FIG. 2, the sampling frequency εE is thus halved compared to the image capture reference method which was recalled above.

The invention is of interest when the offset length M is selected to further verify the inequality M> 1 / (2 ^■ f _c ^■ p). Then, the sampling frequency ÎE is smaller than twice the cutoff frequency fc of the optical 100: f _E <2 ^■ f _c . The invention thus relates to image sampling conditions for which spectrum folding occurs, and is capable of producing artifacts in the sampled image that is captured. The contribution to the modulation transfer function that is due to the image sensor 1, called the second contribution in the general part of the description, and evaluated for the spatial frequency in the image that is equal to half of the sampling frequency, becomes sinc [nL / (2 ^■ M)] by applying the invention. Since M <L <2 ^■ M, this value is strictly between zero and sinc (n / 2) "0.64. In other words, the modulation transfer function evaluated for half of the sampling frequency εE is reduced compared to the reference method where the accumulation signals I of the photodetectors 10 are assigned one-to-one at the points of the image frame. The spectral components of the image that are folded thus have contrast values that are reduced, so that the image that is constructed from intensity values h, h- - - exhibits artifacts due to aliasing. , which are diminished.

For the example with L = 3 photodetectors and M = 2 photodetectors, the contribution to the modulation transfer function that is due to the image sensor 1, at half the sampling frequency ÎE, is equal to sinc ( TT-3/4) ~ 0.30. In practice, the spectral components of the image whose spatial frequencies are between half of the cutoff frequency fc and this cutoff frequency fc itself, have contrast values which are reduced. Indeed, the values of the modulation transfer function are low for these spatial frequencies. Artifacts that are due to aliasing from this range of spatial frequencies are therefore minor, so that the sampling frequency ÎE may be less than the cutoff frequency f _c , beyond being less than twice of f _c , while obtaining a captured image devoid of artifacts that are important. Let: f _E <fc, or M> 1 / (f _c ^■ p). The overlap between two successive segments of photodetectors 10 can therefore be reduced while applying the invention.

For example, the shift length M can be chosen to be 8 / (f _c ^■ p). In this case, the sampling frequency f _E is equal to 1/8 of the cutoff frequency f _c only.

Another advantage results from the increase in the effective radiation collection area, which is associated with each intensity value of the image. This effective collecting surface is multiplied by the number of photodetectors 10 which are grouped by segment, that is to say by the length of segment L. Therefore, the total accumulation signal associated with each segment S1, S2, S3, S4 ..., is multiplied by L, on average, while the photon noise level is multiplied at the same time as by the root-square of L. The signal-to-noise ratio that is associated with each intensity value of the captured image is thus itself increased by a multiplicative factor equal to the square root of L , which is greater than unity. Hence an improvement in the signal-to-noise ratio of the captured image.

By way of comparison, FIG. 3 illustrates a reading mode of the image sensor 1, which proceeds by agglomerating neighboring photodetectors by successive disjoint segments, for example by disjoint pairs of successive photodetectors. In other words, the segment length and the offset length between two successive segments are equal to: M = L, for example both equal to two photodetectors 10. This reading mode therefore does not use the present invention. S1 ', S2', S3 ', S4' ... denote the new segments which are thus defined in the row of photodetectors 10 of the sensor 1, in place of the segments S1, S2, S3, S4. The reading mode sampling of Figure 3 is twice as small as that of the reading mode of Figure 2, and the contribution of the sensor 1 to the modulation transfer function for half of the sampling frequency is again equal to ^" 0.64 for Figure 3, instead of ^" 0.3 for Figure 2. For these reasons, the reading mode of Figure 2, which is in accordance with the present invention, produces less than artifacts in the image that is finally grasped.

On the other hand, the value of the signal-to-noise ratio is also lower for the reading mode of FIG. 2, compared with that of FIG. 3, by a factor equal to (3/2) ^{1 2} "1 22.

However, the two reading modes of Figures 2 and 3 correspond to amounts of image data that are identical.

In polyvalent embodiments of the invention, the two reading modes of the image sensor, respectively in accordance with FIGS. 2 and 3, can be selected alternately at will, depending on the imaging mission concerned. Such a selection can be made by a appropriate command that is addressed to the controller 2, for example.

It is understood that the invention as just described can be adapted or modified in details of its implementation, while retaining at least some of the advantages that have been cited. In particular, the lengths of the segment L and the offset M between two successive segments can be increased, but however the segment length L remains strictly between the offset length M and the double of the latter, for the application of the invention .

Claims

1. Image capture instrument, comprising:

an image forming optics (100) adapted to form an image of a scene in a focal plane (103) from radiation from said scene, the imaging optics having a diameter D of pupil (P) and a focal length f determining a numerical aperture N and an effective cutoff spatial frequency f _c for patterns in the image, according to the formulas:

N = f / D and f _c = 1 / (N ^■ λ), where λ is the radiation wavelength;

a row (1) of photodetectors (10) located in the focal plane (103), with a pitch p of the photodetectors which is constant along the row; and

a controller (2) adapted to control readings of accumulation signals (I) produced respectively by the photodetectors

(10)

an instrument in which the controller (2) is further adapted to define a series of segments (S1, S2, S3 ...) within the array (1), each of which comprises the same number (L) of photodetectors (10) adjacent, called segment length, and to control readings of the accumulation signals (I) by adding said accumulation signals for photodetectors belonging to one of the segments, and the controller (2) is also adapted to define the segments (S1, S2, S3 ...) so that an offset length M between two successive segments in the row (1), expressed as a number of photodetectors (10), is constant along the row, and greater than 1 / (2 ^■ f _c ^■ p), the instrument being characterized in that the segment length (L) is strictly greater than the offset length M, and strictly less than twice said length shift M.

2. The instrument of claim 1, wherein the segment length (L) is equal to three adjacent photodetectors (10), and the offset length M between two successive segments (S1, S2, S3 ...) in the row. (1) is equal to two adjacent photodetectors.

3. Instrument according to claim 1 or 2, wherein the shift length M between two consecutive segments (S1, S2, S3 ...) in the row (1) is greater than 1 / (f _c ^■ p).

4. Instrument according to claim 3, wherein the shift length M between two segments (S1, S2, S3 ...) in the successive row (1) is between 2 / (f _c p ^■) and 10 / (f _c ^■ p).

An instrument according to any one of the preceding claims, wherein the controller (2) is adapted to control readings of the accumulation signals (I) according to a plurality of alternative read modes, the addition of the accumulation signals for the photodetectors (10) belonging to one and the same segment (S1, S2, S3 ...) when the segment length (L) is strictly greater than the offset length M and strictly less than twice said offset length M, corresponding to a first of the reading modes, and the controller (2) is further adapted to, in another of the reading modes, defining another series of segments (S1 ', S2', S3 '...) to inside the row (1) of the photodetectors (10), with the same offset length M between two successive segments as in the first read mode, but with a new segment length which is equal to said offset length M, and to add the signals of accumulation (I) for photodetectors belonging to the same segments of said other series.

An image capture method, comprising the steps of:

providing an imaging optic (100) adapted to form an image of a scene in a focal plane (103) from radiation from said scene, the imaging optics having a pupil diameter D (P) and a focal length f determining a numerical aperture N and an effective cut-off frequency fc for patterns in the image, according to the formulas:

N = f / D and f _c = 1 / (N ^■ λ) where λ is the radiation wavelength; in the focal plane (103), arranging a row (1) of photodetectors

(10) with a pitch p of the photodetectors which is constant along said row;

defining a series of segments (S1, S2, S3 ...) inside the row (1), each of which comprises the same number (L) of adjacent photodetectors (10), called segment length; and

controlling a reading of accumulation signals (I) produced respectively by the photodetectors (10), by adding the accumulation signals of the photodetectors belonging to one and the same segments, an offset length M between two segments (S1, S2) , S3 ...) in the row (1), expressed as a number of photodetectors (10), being constant along the row, and greater than 1 / (2 ^■ f _c ^■ p), the method being characterized in that the segment length (L) is strictly greater than the offset length M, and strictly less than twice that of the offset length M.

7. The method of claim 6, wherein the segment length (L) is equal to three adjacent photodetectors (10), and the offset length M between two successive segments (S1, S2, S3 ...) in the row. (1) is equal to two adjacent photodetectors.

8. The method of claim 6 or 7, wherein the shift length M between two successive segments (S1, S2, S3 ...) in the row (1), is greater than 1 / (f _c ^■ p).

9. Method according to any one of claims 6 to 8, wherein the shift length M between two successive segments (S1, S2, S3 ...) in the row (1), is between 2 / (f _c ^■ p) and 10 / (f _c ^■ p).

10. Method according to any one of claims 6 to 9, wherein the addition of the accumulation signals (I) for the photodetectors (10) belonging to one of the same segments (S1, S2, S3 ...) when the segment length (L) is strictly greater than the offset length M and strictly less than twice said offset length M, corresponds to a first reading mode from among several alternative reading modes, and in another one of reading, another series of segments (S1 \ S2 ', S3' ...) is defined inside the row (1) of the photodetectors (10), with the same lag length M between two successive segments as in the first read mode, but with a new segment length which is equal to said offset length M, and the accumulation signals (I) are summed for photodetectors which belong to one of the segments of said other series, the process comprising a initial step of selecting one of the reading modes.

1 1. A method according to any of claims 6 to 10, wherein the image forming optics (100) and the array (1) of the photodetectors (10) are parts of an image pickup instrument, said instrument is mounted on board a satellite, and the method is implemented to capture at least one image of a surface portion of the Earth.